L0rdGwyn's DIY Audio

[L0rdGwyn]

*I originally created this thread for a single amplifier project. I am planning to pursue others, this thread will now encompass all things DIY audio

Hey Head-Fiers,

I have decided to take on a big project, a full-sized, from scratch, single-ended triode headphone/speaker amp. Prior to this build, I have built/modded after-market Grado-style headphones, built my own headphone cables and interconnects, completed Bottlehead kits, restored a tube tester, and made circuit modifications to my Glenn OTL. I have learned a great deal along the way, but this will be my first time building an entire amplifier.

The amp will be heavily inspired from the SET builds done by @2359glenn and I will be leaning on him for his expert opinion as I move along, he has graciously agreed to help me out! Doing a lot of reading on circuit design and practical building considerations as well.

While the topology is common, the tubes that will be utilized in this amp are not. This will be the first time these tubes have been used in a SET headphone amp, but from my listening tests, I doubt it will be the last. This family of tubes are mostly pre-WWII 4V/1A filament medium to high μ preamp tubes.

Here is the star of the show, the Marconi MH4.



While the input stage will be optimized for the MH4, it has numerous near-equivalents that will be compatible and also sound fantastic, here are a few I have in my collection with more to come:



In terms of outputs, it will utilize EL34 pentodes as well as KT66 and 6L6G beam tetrodes, all run as strapped triodes. There is a possibility of using other tubes as well (KT77, EL33, KT63), still working through that process. Mains and output transformers will be Lundahl.

The power supply design is a work-in-progress, but it will have a c̶e̶n̶t̶e̶r̶-̶t̶a̶p̶p̶p̶e̶d̶ ̶m̶a̶i̶n̶s̶ ̶t̶r̶a̶n̶f̶o̶r̶m̶e̶r̶ ̶f̶o̶r̶ ̶u̶s̶e̶ ̶o̶f̶ ̶a̶ ̶f̶u̶l̶l̶-̶w̶a̶v̶e̶ ̶t̶u̶b̶e̶ ̶r̶e̶c̶t̶i̶f̶i̶e̶r̶.. Update: while building the schematic, I discovered the original planned transformer did not have a sufficient current rating for the output tubes, so a non-center-tapped transformer will be used with a diode-tube hybrid bridge rectifier. The amp will be able to take rectifier tubes of two different bases and filament voltage ratings in a single socket, 5V UX4 (e.g., 5Z3, 80, 274A) and European 4V B4 (e.g., U18/20, RGN2504, 45-IU, NU12).

Here are some of the rectifier and power tubes I have in my collection already, more to be added as the amp is constructed:



Part selection and planning has already been taking place for a month or two. The amp will use teflon Yamamoto sockets, including the dual UX4-B4 rectifier socket. The chassis is currently being built by Landfall Systems, it will be anodized in gold. A diagram of the chassis layout is included below:





Overall, this will be a long process. Currently in the design phase and drafting the schematic while simultaneously doing some part selection. As of right now, I plan to prototype and test the amp before doing the final build in the chassis. This will allow me to make modifications and optimize the circuit. Once I have the test equipment, I will source inexpensive parts for the prototype before purchasing boutique components. Here are some of the parts I am planning to use, some are on hand already:

Goldpoint stepped attenuator
Yamamoto teflon sockets
Lundahl mains and output transformers
Audio Note Standard and Kaisei electrolytic capacitors (power supply, cathode resistor bypass)
Jupiter copper foil coupling capacitors
Kiwame carbon film and Riken carbon composition resistors
WBT-0201 RCA sockets
DH Labs hook up wire (will likely limit this to signal path only, +/- on the idea right now)

I'll leave it there for now and update as I go along! Very excited to be working on this project and planning to set myself up for future amp builds down the road. Tube audio is something I have become very passionate about, time to finally unleash it

 

[L0rdGwyn]

Since I first made this thread three days ago, with no exaggeration, I have been at it sun up to sun down working on the schematic for this amp. I have consumed an unholy amount of information in that time, if I wasn't so excited about this I would never would have been able to keep up the pace.

But I am happy to say the power supply and amplifier schematics are complete. Big thanks again to @2359glenn for signing off on my designs.

I'll just discuss the power supply in this post. This gave me the biggest headache, but I finally feel it is right. Here is the schematic:



Transformer:
The transformer will be a 350V 630mA Lundahl LL1650.

High-voltage B+:
Rectifier will be a hybrid HEXFRED diode-tube bridge. Rectifer tubes will vary, but the bottom line is they will be able to supply around 250mA max to the tube plates. The filter will be a capacitor input C-L-C-L-C design with an additional R-C filter for the driver tubes. The resevoir capacitor will be an Auricap XO+ 2.2uF 600V, this cap is responsible for driving the rest of the supply and gets "topped up" by the rectifier as it is discharged. The two L-C filters following the resevoir cap are responsible for smoothing the rectified DC before it is send out to the plates of the power tubes. The capacitors will all be Audio Note Kaisei electrolytics. The R-C filter before the driver tubes serves the same purpose, but also drops the B+ voltage sent to the plates of the driver tubes by ~20V. An Audio Note Kaisei capacitor will be used here as well.

This supply was iteratively modeled in Power Supply Designer II software, a free tool that is incredibly useful for modeling supplies, used by many amp building hobbyists. With this setup, the model indicates I will be getting 13mV of ripple on the power tube plates and 0.34mV of ripple on the driver tube plates. Capacitor values may change in the final design after prototyping, so the ripple may go up or down a bit.

Here is the circuit in the model and a graph of the output voltage at B+ and B++. Note that the rectifier is a full-wave tube with a center-tapped transformer. The software cannot do a diode-tube hybrid rectifier, but the end result is the same.





Filaments:
The power tubes will get 6.6VAC to the filaments without any fancy filtering or regulation.

Because the MH4 driver tubes take 4V 1A to the filaments, the incoming 6.6VAC has to be dropped. In addition, for the quietest possible input stage, the incoming current will be rectified and smoothed to DC. After going over a few different solutions, I decided I am going to use Pete Millett's (a well known tube DIYer) regulated DC filament supply, here:



http://www.pmillett.com/DC_filament_supply.htm

This board was originally designed and intended for use for directly-heated triode tubes (i.e., tubes where the filament is directly connected to the cathode, so any AC hum can be heard on in the output signal), which require DC on the filaments for audio use. The driver tubes for my amp are indirectly heated, so not prone to the same hum of directly-heated triodes, but it will serve the same purpose, a silent filament supply.

Total cost for the printed boards from his eBay shop and the parts from Mouser is $80. The circuit includes a variable resistor to fine tune the output voltage to the needs of the tube filaments.

Think that's it for the power supply, now on to the amp circuit!

*Update 1/7/2020: through iterative reevaluations of the amplifier design, the power supply has been altered, please see this post for the most recent build, the above information is no longer up-to-date. Also, the amplifier will no longer feature direct current on the driver tube heaters.

https://www.head-fi.org/threads/l0rdgwyns-single-ended-triode-diy-amp.921105/page-5#post-15403852

 

[L0rdGwyn]

Now the amplifier circuit. Here is the schematic:



Potentiometer:
This will be a 25K Goldpoint stepped attenuator. The volume pot acts as a voltage divider between the driver tube grid and ground. That means turning up the volume pot will increase the resistance between the input signal and ground, so more signal is sent to the driver tube grid, resulting in increased signal and volume. Turning down the volume pot decreases the resistance between the input signal and ground, so more of your input signal goes to ground and your volume decreases. With the volume set to 0, all of the signal is lost to ground.

Input stage:
It will feature my new favorite tube, the MH4 and its cousins. It will be cathode biased, also known as auto-biased. Without getting into the nitty gritty details, this will alow the tube to find its own bias point based on the value of the cathode bias resistor and allows the tube to maintain that bias point as it ages. This biasing scheme requires a cathode resistor bypass capacitor, which acts as a high-pass filter, among other things. The value is appropriately high to allow all audible low frequencies to pass. This will be an Audio Note Kaisei electrolytic. The 56kohm resistor on the plate is known as, you guessed it, the plate resistor this acts as the load for the tube and is usually 2-3 times the internal resistance of the tube. From experimenting with GOTL modifications, I have found that changing this resistor has a very audible effect on the sound. I will be using discontinued Japanese Riken carbon composition resistors on the MH4 plates, considered by many to be the best sounding resistors made.

I am going to briefly explain how the bias point is chosen for the input tube using "load lines". Below is the current vs. voltage operating characteristics of the tube with the load line drawn in red. The far right point of the line represents the maximum voltage across the tube, equal to the high voltage B++. The far left line point represents the current flowing across the 56kohm plate resistor with zero voltage across the tube. This load line tells us the plate current for any given voltage on the tube and our bias point will be on this line. The curved lines on the graph represent the grid voltage, and the grid voltage at our desired bias point will determine the value of our cathode bias resistor.

For the input tube, we want to choose a linear point on the load line, meaning we want to choose a point on the load line where the distance to the next nearest grid voltage line on both sides is about equal. The tube will swing along the line right to left with the positive and negative AC input signal.

I chose to bias the MH4 with -3V on the grid, which corresponds with 145V on the plate, a plate current of about 1.75mA, and a cathode bias resistor of 2kohm. This is the green dot on the load line. I may adjust this in the final build to a grid voltage of -2.5 or -2 depending on the distortion measurements.



The input stage is coupled to the output stage via a Jupiter copper foil 0.47uF 630VC capacitor. This allows only the AC output signal to pass and blocks the 145V DC from the MH4 plate. If that DC current reached the grid of the next tube, it would "red plate" and burn up!!!

*Update 1/7/2020: in the final design, the driver tubes will now be cascode constant current source (CCS) loaded rather than using a resistive load. The method of drawing load lines described above only pertains to using a resistive load, not a CCS.

Output stage:
The output stage of this amp will feature various pentode and beam tetrode power tubes as strapped triodes. This includes EL34, KT66, 6L6G, KT77, and the possibility of others. A Goldpoint selector switch will be used to change the cathode bias resistor for the optimal bias point for each tube. The switch has six possible positions, so I can add resistors for two more tubes later if I please in reality, these tubes could probably all be used with the bias optimized for one tube and no switch. Rarely are headphones going to require enough power to push the tubes to significant distortion at an unoptimal bias point, this is much more important for higher power uses like speakers.

But screw that!!! I will optimize the bias of each tube and choose the appropriate resistor. Audio Note Kaisei will be used again for the bypass capacitor. For strapped triode mode, the screen grid is connected to the plate via a 100ohm resistor and the suppressor grid is connected directly to the cathode (in beam tetrodes, this is an internal connection as there is no true suppressor grid, the suppressor grid is actually generated by the beam of electrons flowing through the tube!!! Very cool, look it up).

Using our load lines again, we can choose the bias point for the power tubes. However, it is a bit of a different process since we do not have a plate resistor acting as the load for the tube, the AC load is the impedance of the primary winding of the output transformer, 4.6kohm. Also, the primary winding has a very low DC resistance, so the plate of the power tube essentially sees the entire 250VDC B+.

Like the input tube, we can draw our load line with the full B+ voltage on the right and the maximum plate current on the left, the blue line on the graph. But our true operating point is not on this line! We cannot choose the plate voltage since there is no plate resistor, the plate voltage is FIXED as the B+ of 250V. If we operated the tube here, the bias point would be the far right of the load line, there would be no place for half of our input signal to go and it would be lost.

So, we have to draw additional load lines at the same slope as our initial line, up to but not exceeding the large curve at the top of the graph. This is the maximum plate dissipation of the tube. If we exceed this line with our operating point on the load line, the tube will melt. We want to choose a load line that is below the maximum plate dissipation and allows for maximal left and right excursion along the load line. The upper limit on the right side of the line this time around is where the grid curves start to get squished, which will cause significant 2nd harmonic distortion. The leftmost limit is the grid voltage where the grid will begin to draw current, which is BAD. By maximizing the equal left to right swing across the load line, which are getting the most peak-to-peak voltage out of the power stage, which means more POWER!

I will have to redo my bias points since I made changes to the power supply, but for the sake of example, here is where I might choose the bias point of the EL34, green dot on the red load line. This point represents 250V on the plate, 85mA plate current, a grid voltage of about -14V, and a cathode bias resistor value of ~165ohm. This will be done for each power tube to find the correct cathode bias resistor value for the switch.

Doing some quick math, at this bias point, this gives us roughly a peak-to-peak voltage of 220V, which is about 77VRMS, which means about 1.3W of undistorted output power. If I were to push the output tubes closer to the max plate dissipation, could get something like 2.5W out of them with this output transformer. Either way, it is way overkill. The tubes will live a longer life at this bias point and a power output of 1.3W.



Okay, wrapping this up, the output transformer will be a Lundahl LL2765 with a 4.6kohm primary and a 32ohm secondary winding, which will be the output impedance of the amp.

Well there you have it, the amplifier circuit. I know that is technically dense, but hopefully somewhat interesting.

Next step will be building the parts list and prototyping this bad boy!!!

BTW, this is still new to me, so please if Glenn or anyone else spots an error feel free to let me know

*Update 1/7/2020: for the sake of optimization and other design considerations, the output stage of the amplifier will no longer use multiple power tubes. In fact, they have been changed altogether to the 6A5G tube, although the EL34 is a possible alternate depending on the final sound of the amp. The above information on the output tubes is no longer up-to-date, although the method for determining the load line of an SET output stage is still relevant. Please see this post for more information on the current power tube, the 6A5G.

https://www.head-fi.org/threads/l0rdgwyns-single-ended-triode-diy-amp.921105/page-5#post-15403852

 

[maxpudding]

This is exciting! Best of luck LG. Gonna follow your thread to learn more about amps

 

[chrisdrop]

This is exciting! Best of luck LG. Gonna follow your thread to learn more about amps

.. what he said .
Me too!

 

[BunnyNamedCraig]

Very cool!!!

excited to see the progress.

Subbed btw

 

[L0rdGwyn]

Very cool!!!

excited to see the progress.

Subbed btw

Thanks! I have been at it non-stop for days, things are coming along nicely, will have the schematic complete probably by the end of the day today, it has been approved Glenn, but I am going to make a few tweaks to the power supply. Once it is complete, the next step will be to prototype the power supply and measure it.

Will post some updates in my reserved posts above later today, hope to make them somewhat educational on choosing bias points, tube linearity, etc. but at a high level, no maths

 

[BunnyNamedCraig]

Thanks! I have been at it non-stop for days, things are coming along nicely, will have the schematic complete probably by the end of the day today, it has been approved Glenn, but I am going to make a few tweaks to the power supply. Once it is complete, the next step will be to prototype the power supply and measure it.

Will post some updates in my reserved posts above later today, hope to make them somewhat educational on choosing bias points, tube linearity, etc. but at a high level, no maths

Nice! I hate math!!!

 

[L0rdGwyn]

Hey people - just updated my posts above with schematics and explanation of the circuit. I learned this over the past three days, non-stop, has been a blast but my brain is mush.

Sort of techncially dense but I hope it interests people! Next I will be getting parts together to prototype and test the power supply.

 

[L0rdGwyn]

One more thing, got these beautiful Philips E424N tubes in the mail yesterday. They are MH4 near equivalents and will be used as drivers in the amp, similar to what Glenn is using in his preamp.

 

[SonicTrance]

Hey people - just updated my posts above with schematics and explanation of the circuit. I learned this over the past three days, non-stop, has been a blast but my brain is mush.

Sort of techncially dense but I hope it interests people! Next I will be getting parts together to prototype and test the power supply.

Congrats on getting in to the hobby of building amps! It's very fun indeed!

Schems look good. I was wondering about the high W grid stoppers though? There's no need for more than 1W, even lower is fine. I recommend to learn LT spice when you have many hours left over, lol. It's a great tool to use and it's free.

I'd run those MH4 tubes on at least 3 mA, preferably 4 mA, to get over the "knee" in the grid curves. There you have the best linearity. So it'll sound the best

 

[L0rdGwyn]

Congrats on getting in to the hobby of building amps! It's very fun indeed!

Schems look good. I was wondering about the high W grid stoppers though? There's no need for more than 1W, even lower is fine. I recommend to learn LT spice when you have many hours left over, lol. It's a great tool to use and it's free.

I'd run those MH4 tubes on at least 3 mA, preferably 4 mA, to get over the "knee" in the grid curves. There you have the best linearity. So it'll sound the best

Hey thanks! The high wattage of the grid stoppers is unnecessary, it's only because I want to use a specific brand of resistor and they are only made in 2W and 5W just one of those neurotic things.

I started messing around with LTSpice the other day, but once I got started I realized the learning curve was going to require a lot of time, so I shelved it. Definitely going to try it again when I have time!

And I agree with you on the MH4, I was actually thinking about this earlier today. Will probably increase the value of the resevoir cap and/or drop the impedance on the R-C filter to get a higher B+ to work with. I have a bunch of different cap and resistor values on my shopping list so I can optimize all this stuff when I build out the power supply.

Edit: wait I take it back, forgot the stoppers need to be non-inductive. I'm gonna change them to 1/2W carbon comps.

 

[2359glenn]

Yes the grid stoppers should be carbon composition. No real current going through them so won't be making noise

 

[L0rdGwyn]

I think I might have to look into having some custom chokes wound from Heyboer @2359glenn . To get the MH4 up to >3mA on the plate like @SonicTrance suggested, need more B+ and I'm losing too much voltage on the chokes with the current draw of the power tubes. Could go to 4H 65ohm 225mA but not great losing the filtering, think I end up with 14mV ripple on the power tubes.

Running everything at a higher voltage will increase the linearity of the MH4, but the linearity of the EL34 bias point will suffer. Going to have to find a balance, but will favor the MH4.

Man, you change one thing in this business and everything else changes too! Tricky tricky.

Edit: I might have to bite the bullett and lower my plate resistor. Was hoping to keep 56kohm since I have these Riken resistors, but the tubes will be biased better at 30-40kohm.

 

[mordy]

HiLG,
Very interesting to read about all the particulars in building your amp, but not having any electronic background I can't say that I understand much about what you are describing.
Apart from the tubes you are using, is this a conventional type tube amp or an innovative design?